the Air Vent

Because the world needs another opinion

Alaska Bodge Answers

Posted by Jeff Id on December 12, 2009

Well, I’d like to thank Dr. Keen for stopping by and explaining some of our questions. This thread which pertains to the Alaska temperature record.

Those few Alaska graphs can sure stir up a lot of conversation! I posted them on ICECAP as an illustration of how original station data can differ from “processed” or “value added” data released by various agencies.
For those who wonder what the original data came from, I went straight to the source, of course, namely NCDC collection of co-op and NWS station observations at:
http://www.ncdc.noaa.gov/oa/climate/climatedata.html#monthly
You, too, can download the same data, but it might cost you.
The nine stations are the only long-term ones in the GHCN grid box that’s shown; I have no idea what area the IPCC used. Seems like the IPCC may not know either.
As for my secret code, you can dowload that too – just open Excel and find the AVERAGE, STDEV, COUNT, and SUM funcions. That’s as fancy as my codes got. I have no fudge factors, except for arithmetic offsets between overlapping stations that were used to make combined time series. Some time series were normalized (departure divided by standard deviation), and some weren’t. The full report is a 70 or so page tech report for the National Park Service. A nice summary of the work (and the purpose of the study) was published by NPS at:
http://www.nps.gov/akso/AKParkScience/symposium2006/sousanes.pdf
The lesson I learned from this is that one should use original source data (like NCDC co-op data) rather than processed, or possibly processed (sometimes you don’t know) data from sources like USHCN, GHCN, GISS (not an issue with CRU, since they won’t release it).
I haven’t visited this blog before, and although it looks pretty good, I probably won’t be checking back to reply very often. Nothing wrong, mind you, it’s just a matter of available time.
Meanwhile, I need to go read my official NWS issue MMTS max-min thermometer now – I’m the co-op observer for Coal Creek Canyon, Colorado, elevation 8950 feet. Check it out – we get some cool numbers here. Such as, every one of the past five years has been colder than the coldest of the previous five years.
cheers, Richard Keen
Climate analyst and data maker (observer)

43 Responses to “Alaska Bodge Answers”

  1. windansea said

    Such as, every one of the past five years has been colder than the coldest of the previous five years.

    cool!

  2. JWDougherty said

    I’ve been thinking that the “fudge factor” in Harry’s code puts an entirely new cast on the words “value added.” That’s likely why there are a lot of data sets out there that the curators make it difficult to access.

  3. VG said

    Ot but Willis needs to deconstruct this from the economist regarding Darwin temps
    http://www.economist.com/blogs/democracyinamerica/2009/12/trust_scientists

  4. stan said

    VG,

    The argument in the Economist is basically that scientists are way, way smarter than we are about statistics, so we need to shut up and trust whatever they tell us. Of course, Professor Wegman (one of the world’s foremost stats experts) wrote that climate scientists would produce better quality work if they consulted with stats experts and it was a shame that they refuse to do so.

  5. Nick Stokes said

    Willis’ Darwin story seems to have fallen apart. Giorgio Gilestro says he has done the calculation that should have been done in the first place. Instead of picking out one station, he has looked at all the stations in the GHCN set for which adjustments were performed. He shows the distribution of the effects of the adjustment on trend. It looks pretty symmetric. Adjustments are just as likely to cool as to heat. And his Python code is available.

  6. V said

    I think the Giorgio Gilestro post mentioned above is not relevant. He purports to show that as the distribution of adjustments is symmetrical. However, this is not enough. One could produce a warming trend by only using negative adjustments. It the temporal arrangement of the adjustments that is important.

  7. Nick Stokes said

    V #6 No, the plot Giorgio shows is the distribution of trend increments due to adjustment – deg C /decade.

  8. Lady in Red said

    Apparently, consensus has been achieved: no problem, just boys!

    http://news.yahoo.com/s/ap/20091212/ap_on_sc/climate_e_mails

    BY SETH BORENSTEIN, RAPHAEL SATTER and MALCOLM RITTER, Associated Press Writers – 1 min ago
    LONDON

    Associated Press writers Jeff Donn in Boston, Justin Pritchard in Los Angeles contributed to this report. Troy Thibodeaux in Washington provided technical assistance. Satter reported from London, Borenstein from Washington and Ritter from New York.

  9. Ryan O said

    Nick, that has nothing to do with what happened at Darwin. The Darwin story hasn’t fallen apart – it is still as valid as ever. Even if the average adjustment is zero, doing stuff like what is done at Darwin affects the spatial distribution of temperature change, so it’s bad regardless. And I note that even Giorgio’s calculations yield a 0.017 C/decade – or 0.17 C/century – warming trend, and given a supposed underlying trend of 1.2 C/century, comprises 1/6th of the trend. Not only that, but the stations are, of course, unevenly distributed geographically, so a simple histogram of trend adjustments doesn’t tell the whole story. Until you grid it up, you have no idea how those adjustments change the global trend.

    But you keep those blinders on if it makes you feel better.

  10. Paul Dennis said

    Rather than he temporal arrangement it is the spatial arrangement of the adjustments that is important. The ‘average’ global temperature is computed from an inhomogeneous distribution of stations with a highly variable sampling density. A normal distribution of adjustments across all the stations centred on zero can still propagate through to an increase in global average temperature.

    What is concerning is that the adjustments are made without any concern for ‘physical reality’. At one point, and for an extended period there were 4 stations in and around Darwin all recording very similar temperatures. Presumeably each of these sites had different characteristics yet each agreed with the other. Given this any empirical scientist would find it hard to justify any adjustment. So we are left with adjustments being made by an algorithm without concern for the physical reality.

  11. Paul Dennis said

    Sorry Ryan..I should have refreshed. I went straight to Giorgios analysis and then commented there and here. Snap we agree completely.

  12. Creation Man said

    Sounds as though the globe might be cooling, not warming.

    WattsUpWithThat has something to say about that, too, and so did Joe Bastardi of AccuWeather. Here’s the full treatment:

    http://www.examiner.com/x-28973-Essex-County-Conservative-Examiner~y2009m12d12-Globe-might-actually-be-cooling

  13. anon said

    NCDC Coop data is NOT, repeat NOT raw data. They “correct” it.

  14. Dribble said

    Paul Dennis: “What is concerning is that the adjustments are made without any concern for ‘physical reality’. At one point, and for an extended period there were 4 stations in and around Darwin all recording very similar temperatures. ”

    I’m completely confused by the mystery of the five GHCN stations at Darwin Airport mentioned in Willis’ article. The five records appear to be a composite of only two stations. The first is Darwin Post Office whose record is used up to 1942, then Darwin Airport whose record is used from 1942 up until 2009.

    According the Bureau’s list, the stations for Darwin are as follows:

    014015 DARWIN AIRPORT Jan 1941 Aug 2009
    014040 DARWIN AIRPORT COMPARISON Aug 2001 Jun 2007
    014016 DARWIN POST OFFICE Jan 1885 Jan 1942
    014161 DARWIN REGIONAL OFFICE Jan 1967 Dec 1973

    Darwin Airport Comparison and Darwin Regional Office don’t seem to have anything to do with it. It seems to me that the GHCN bureaucracy has sprouted more stations than are actually there.

  15. Genghis said

    Nick Can you show in Giorgio’s code where the warming adjustments equal the cooling adjustments in the 30’s and 40’s like the histogram suggests. I think that is what Giorgio is implying anyway.

    If the ‘climatologists’ adjust the temps upward from 1960 onward and downward for the pre 1960’s you will get the same histogram.

  16. Kenneth Fritsch said

    Ryan O, makes an important point above in reply to Nick Stokes. I see that all sides of the AGW issue and the proxy and thermometer issues sometimes miss the point.

    Some seem not interested in any evidence of proxy/thermometer error/uncertainty that does not “prove” that we have had no or little warming in the modern instrumental era or worse misinterpret what uncertainty means, i.e. the truth could go either way. And others seem to want to point to aggregate results such as global temperatures versus regional ones and periods of agreement and not disagreement between temperature data sets. They emphasize the symmetry of the adjustments, without addressing the size of the adjustments and what that might imply for the uncertainty of measurements.

    I think the true scientist, in cases like this one, would want to disaggregate and look for uncertainties in the measurements and particularly so when we have no absolute standard for comparisons. I think, most naturally and expectedly, that the advocate who likes the current consensus, for whatever reasons and motivations, is going to want to avoid looking at those aspects of the instrumental and proxy records that might show uncertainty.

    I plan to show more difference series for temperatures data sets (this time land/surface) for zonal regions of the globe and point to the difference series’ breakpoints that can be visualized (objective measures to come later) and the statistical significance of differences which are large.

    I think when claims are made for given temperature data sets error bars and then one can show, in disaggregate, were those limits are exceeded when compared with another data set with its own claims one can at least make a claim against the uncertainties claimed or implied by the data set owners.

  17. Todd G said

    To obtain what seems to be unprocessed (unhomogenized, unpasteurized) ‘free’ data for western US stations try http://www.wrcc.dri.edu/, the Western Region Climate Center in Reno NV. Other unhomogenized ‘free’ data may be available from the other regional data centers… see http://www.wrcc.dri.edu/rcc.html for links.

    Reading (between) the webpage lines, WRCC’s data sources are NOAA plus state climate offices.

    A quick comparison of eastern CO station data from WRCC to USHCN or GISS data shows substantial differences. For example, at Holly CO a decline in linear temp trend in WRCC data from 1895-2008 (-1.7 degF/century) turns into an increasing trend (+1.5 degF/century) for USHCN.

    For an average of WRCC temperatures at 6 eastern CO stations, WRCC data show a slight positive trend of 0.4 degF/century since 1895, and 3.5 degF/century since 1981. GISS at 103W, 39N (eastern CO) shows 5.6 degF/century since 1981; the difference of recent (post-1980) trends between GISS and WRCC is 2.1 degF/century…

    We need to remember that USHCN is one of the inputs to GISS, according to http://chiefio.wordpress.com/gistemp/ (see 22 Jul 09 comment). Hence, biases or errors in NOAA’s ‘homogenized’ data will be reflected in GISS outputs; error or bias propagation is an issue.

    To get to the crux of the temperature trend ‘divergences’, FOIA requests need to go to NOAA for their program codes used to generate USHCN temperature values. Some documentation of NOAA’s algorithms and results are available on the web. I’ve not been able to find NOAA code, but I’ve not pressed the search.

    In the meantime, back to the basics: use observed data to try to identify and infer temperature trends… results from Darwin, Alaska and now Colorado suggest that trends from observed data are substantially different than those computed from ‘homogenized, value-added’ data sets.

    In which other geographic areas do these differences occur?? Results from an army of number-crunching Davids could define where and magnitude.

  18. Nick Stokes said

    Ryan O #9
    What falls apart is Willis’ underlined claim that

    Those, dear friends, are the clumsy fingerprints of someone messing with the data Egyptian style … they are indisputable evidence that the “homogenized” data has been changed to fit someone’s preconceptions about whether the earth is warming.

    Yes, the GHCN adjustments may affect the spatial distribution of trends. That seems to be the intention. In their overview paperthey neatly anticipate the Darwin story:

    However, on scales of half a continent or smaller, the homogeneity adjustments can have an impact. On an individual time series, the effects of the adjustments can be enormous.

    Therefore, the best use for homogeneity-adjusted data is regional analyses of long-term climate trends (Easterling et al. 1996b). Though the homogeneity-adjusted data are more reliable for long-term trend analysis, the original data are also available in GHCN and may be preferred for most other uses given the higher density of the network.

    There seems to be some confusion about the five different datasets (Darwin Zero ,,,). They aren’t necessarily five different sites. Some may be different record sequences from the same data source. Here ia a comment from a scientist (Blair) who has been involved in adjustments to the Australian data sets:

    In the specific case of Darwin, while we haven’t done the updated analysis yet, I am expecting to find that the required adjustment between the PO and the airport is greater in the dry season than the wet season, and greater on cool nights than on warm ones. The reason for this is fairly simple – in the dry season Darwin is in more or less permanent southeasterlies, and because the old PO was on the end of a peninsula southeasterlies pass over water (which on dry-season nights is warmer than land) before reaching the site. This is fairly obvious from even a cursory glance at the data – the record low at the airport is 10.4, at the PO 13.4.

    Darwin is quite a difficult record to work with. There were 12 months of overlap between the PO and the airport, but the observing site at the PO deteriorated quite badly in what turned out to be its last few years because of tree growth overhanging the instruments. Fortunately, we recently uncovered some previously undiscovered data from 1935-42 from the old Darwin Airport (at Parap) and should be able to use this to bypass the promblematic last few years at the PO.

    The post-1941 adjustments (all small) at Darwin Airport relate to a number of site moves within the airport boundary. These days it’s on the opposite side to the terminal, not too far from the Stuart Highway.

  19. Chas said

    I admit I am a little confused; giorgio seems to be suggesting that the adjustments are roughly symmetrical.
    CRU and GISS to show the same overall temperature trends (using roughly the same temp stations) BUT Phil Jones doesn’t seem to be making this neutrality claim at all—- In slide 8 of the following presentation he shows how dramatic his adjustments have been:

    http://www.cgd.ucar.edu/cas/symposium/061909presentations/Jones_Boulder_june2009.ppt

    Much more in line with what Willis has been suggesting! (?)

  20. Ryan O said

    #18 Nick, so to bolster your claim you:

    1. Quote a paper that admits the individual series may be wildly off (but that’s okay, though, ’cause even if the little picture is all f&*!d up, the big picture isn’t – WTF kind of arm-waving is that???),

    2. Explain that the metadata for the station is AFU and no one is sure exactly what happened at the stations (but somehow that makes it all better . . . WTF???), and,

    3. You quote a scientist who admits he hasn’t done an analysis yet.

    Good show. Yet again, you prove you have nothing of value to add.

  21. Ryan O said

    #18 Also, I note how you move the goalposts (yet again) when shown that your claim is not supported by your evidence. How on earth does your post in #18 relate to ANYTHING you posted in #5?

    Anyway, I will respond when you have something of value to say (which basically means I no longer care what you have to say on this issue).

  22. Nick Stokes said

    No, Ryan, you totally misquote:

    1. Quote a paper that admits the individual series may be wildly off (but that’s okay, though, ’cause even if the little picture is all f&*!d up, the big picture isn’t – WTF kind of arm-waving is that???),
    Not at all true. It didn’t admit that the series may be wildly off – it merely says the adjustments to individual series may be large.

    2. Explain that the metadata for the station is AFU and no one is sure exactly what happened at the stations (but somehow that makes it all better . . . WTF???), and,
    No, he made no complaint about the metadata – he said that the station itself had problems in the 1930’s due to tree growth. And what may make it better is alternative data.

    3. You quote a scientist who admits he hasn’t done an analysis yet.
    No, he says he hasn’t done the updated analysis yet. I quoted him because he, unlike Willis, has knowledge of why the adjustments were actually made. But apparently you don’t want to know.

  23. Kenneth Fritsch said

    Ryan O, I think we should be thanking Nick S for his contributions because what he notes from the literature is exactly what I have noted also. GISS, I know, wants very much for the people using their data to use it at the most aggregated levels. What they fail to note, in my judgment, is that when you take uncertain data at the local level (climate is local for the folks who must vote for AGW mitigation) and average and blend it for regional/global purposes that uncertainty carries forward and is not averaged out like the mean might be. I have actually been contrasting differences between temperature data sets at the regional and global levels and, by golly, you can see significant differences there. Throw in specific (and critical) time periods and you see bigger differences.

    Read my first post and note where I say those people content with the status quo will point. Also note that to the spokespeople for immediate AGW mitigation that the important point is global climate change and the tipping points that change points to in their models. Please remember also that climate models lose much reproducibility when going to smaller subdivisions of the globe. It’s a difference in a frame of mind and way of thinking that we must understand.

  24. Ryan O said

    Nick, no. You are the one who doesn’t understand. On #1, what they are saying is that the adjustments to the individual series can be enormous and therefore primarily useful for long-term regional studies – meaning they do not accurately reflect the local temperature record. If you want something more representative of the local temperature, use the unadjusted series. That is what they are saying. Locally inaccurate, but that’s okay because the adjustments give more accurate “regional” results and are better for “long term” trend analysis.

    Excuse me while I vomit. And read more carefully.

    On #2, they recently recover some previously unknown data (after a problem is pointed out, of course) from a slightly different location from 60 years ago. Yep, that gives me confidence that the history of the station location is complete. Right. You bet.

    On #3, that’s frickin arm waving. The explanation he gives explains the adjustment in ~1941 which Willis AGREES with. It’s the post-1941 adjustments that are the problem – the very one your scientist simply arm-waves away. No one’s arguing about the frickin PO/airport overlap, Nick. What Willis is arguing about are the “small” post-1941 changes that have a cumulative effect that is about equal to the 1941 switch from PO to the airport. If you’d bothered to read Willis’ analysis at all, you’d have understood that. What your scientist is talking about has NOTHING FRICKIN TO DO WITH THE PROBLEMS THAT WILLIS NOTED.

    Nada.

    Zilch.

    Ken, you want to read Nick more carefully, too.

  25. timetochooseagain said

    We should get Christy on this. Considering the excellent work he did in East Africa, Central California, and Alabama with the station data-doing some serious digging and meticulous examination of the records-I’d like to see what would be found in Alaska.

  26. Ryan O said

    By the way, my frustration is because this is Nick’s MO – repeated everywhere:

    1. Find something peripherally related to the question (but something that doesn’t actually address the original question) – like Giorgio’s analysis vs. Darwin.

    2. Make some claim that it refutes something it doesn’t.

    3. When called on it, drop it and submit something else that also is peripherally related, but doesn’t actually address the original question.

    4. Make the same claim that it refutes something it doesn’t.

    5. Repeat as often as necessary.

    His contribution is anything but meaningful.

  27. Nick Stokes said

    Well, Ryan, more on what those quotes actually said:
    meaning they do not accurately reflect the local temperature record. If you want something more representative of the local temperature, use the unadjusted series.
    No, they didn’t recommend the unadjusted series because it was more accurate. They said “may be preferred for most other uses given the higher density of the network”. You need a long contiguous sequence to perform the adjustment, so the adjusted series omits a lot of stations.
    They also said, in the sentence preceding “However”, that the adjustment does have much effect on the global trend, which GG echoes. That’s why they focussed on regional, a scale at which the number of stations is still sufficient, but homogenisation may make a difference.

    On #2, no, there’s no doubt of the history of the station location. It was at the PO until 1941 (just before the bombing), and then it moved to what was then the military airfield at Marara. It seems it wasn’t in the PO grounds, but in the park area across the Esplanade – a prominent location near the cliff edge. Later it moved around at Marara, which postwar became the civil airport as well. Parap was the earlier civil airport, which kept separate records. It was abandoned in the ’50s and subsequently built over.

    On #3, he doesn’t arm-wave the later adjustments away – he says they were due to moving the equipment around the site. The Marara airport developed in major ways. First was the building of a civil terminal post ’50s. Then came cyclone Tracy in 1974, which moved everything. There were several major terminal upgrades.

    But on relevance – Giorgio’s analysis is exactly on point. A reminder again of Willis’ underlined statement:“indisputable evidence that the “homogenized” data has been changed to fit someone’s preconceptions about whether the earth is warming”. GG has shown that applying the same analysis to the whole data set shows that Darwin is way out in the tail of a symmetric distribution. Willis’ evidence looks a whole lot more disputable.

  28. RomanM said

    I wrote up a short post giving some further analysis on GHCN adjustments at

    http://statpad.wordpress.com/2009/12/12/ghcn-and-adjustment-trends/

    amd linked to it to it on his blog. It clearly shows a temporal pattern to the adjustments. his points are not justified.

  29. Nick Stokes said

    Roman, Giorgio’s general response on his blog was to point out that he wasn’t showing a histogram of adjustments to station temperature, but of the resulting effects on the trends (x axis in C/decade), so the temporal issue is less relevant. Have you calculated the trend effects?

  30. RomanM said

    Nick, no, I did not have the opportunity to do such calculation.

    However, what I have shown IS basically the trend amounts due to the adjustments.

    If you look at the matrix math for a regression, you will notice the following. If I regress two variables Y1 and Y2 individually on a predictor X (for the exact same set of sample X values), the slope of the regression of Y1 + Y2 (or Y1 – Y2) on X will be equal to the sum )or difference) of the slopes for the separate regressions. This will vary somewhat if some of the specific X’s are present for one Y but not the other.

    Because of the very large sample sizes involved here, it basically says that the difference between the slopes for the adjusted values and the raw data can be calculated pretty closely from what you see in the second graph in my post because of this additivity property.

  31. At some point, if you build a car from scratch and then the car doesn’t run–and you’ve tinkered with many parts and sub-systems—you have to tear the whole thing down and start again. You don’t want to, but you have to.

    Examine all the raw data from all the stations and all the measuring instruments—air, land, and sea. Can’t get some data because they’ve been destroyed or were never there? Eliminate those instruments/stations. Some data have been altered? Eliminate them. Some instruments turn out to be unreliable? Explain why, and throw out the data they spawned.

    What’s left is a set of usable data.

    Now, propose a model or system for integrating this set of data, a system that will produce a global historical trend for temperature. EXPLAIN WHY THIS SYSTEM IS HIGHLY USEFUL AND NOT MISLEADING. SUBJECT THIS SYSTEM TO WIDESPREAD INSPECTION AND CRITICISM.

    When you’re satisfied the system is useful, plug in the data set and obtain the trend.

    However oversimplified this step-process may be, it at least presents an outline of what should be done. And what should have been done that wasn’t.

    This whole step-process has been carried out in the field of disease causation. It yields very interesting results. For example, one finds the accepted traditional model hasn’t been used, and yet researchers claim they’ve found the cause of a disease. In essence, they have no model, no procedure. Also, the accepted model itself has obvious problems.

  32. Kenneth Fritsch said

    On #3, he doesn’t arm-wave the later adjustments away – he says they were due to moving the equipment around the site. The Marara airport developed in major ways. First was the building of a civil terminal post ’50s. Then came cyclone Tracy in 1974, which moved everything. There were several major terminal upgrades.

    Nick Stokes, can you link us to the documentation of these changes, what changed, when they occurred and whether they correspond with the step changes in raw data and can reasonably explain them?

    Ken, you want to read Nick more carefully, too.

    Ryan O, my only point with the NS post was about the seeming intent of GISS, and I believe also GHCN, in dismissing local (and importantly adjusted) temperatures for use by researchers and instead using the smear called regional and global temperatures. What I do not follow is the logic of getting something more certain by averaging very uncertain local temperatures unless you understand well the uncertainty and that it is completely random. Even random large errors adds to uncertainty depending on sample size avaialable.

    That there are large corrections to individual stations is well recognized by anyone who has worked with them. I showed a while back at CA that the Version 1 and Version 2 USHCN temperatures series have very large differences between some individual stations. Remember we are talking about the same raw unadjusted data and simply different (and only slightly different if one accepts the implications of the authors) adjustments. That the USHCN temperature trend in the lower 48 states over the past century is mostly due to adjustments is also well known and established.

    I do not need a pissing contest to establish any what I have observed and calculated – although I could use help in establishing what it means in quantitative terms to measurement uncertainty.

  33. Ryan O said

    Nick . . . you are being deliberately obtuse. Yes, if you need more stations, you would use the unadjusted. You would also use the unadjusted for the reason I stated. What . . . is there only one singular reason to use unadjusted? Give me a break. If the adjustments (and here I’m talking homogenization, not adjustments due to combining series) cause massive changes to a particular series then that series is no longer representative of local temperature. Local temperature is what the damn thermometer read. Local temperature is the original data. Perhaps you ought to read more of Briggs’ blog: http://wmbriggs.com/

    #2 . . . what? How does that have anything to do with the new station that was just “found” after Willis started questioning the adjustments? Non sequitur. If Darwin was already known to be such a problem site, why did it take 60 frickin years to find that data? I’ll tell you why. Because the problem isn’t with the overlap – the problem is the adjustments at the airport. Willis doesn’t have a problem with the overlap; I don’t have a problem with the overlap; as far as I know, no one has a problem with the overlap. The problem is after the overlap. I note that you ignored this point after I brought it up – par for the course. Your scientist is doing what you do so well – throw in unrelated crap to obfuscate. Maybe that’s why you consider him such a reliable source. Like-minded, eh?

    #3 . . . “due to moving around the site” is frickin arm waving (it’s the same damn airport – how can you get over 1 deg C of adjustments – almost all in the direction of inducing a warming trend – at the same damn airport?). He characterizes them as “small” – which you apparently never bothered to question – because they amount to over 1 deg in about 50 years. 2C per century. But I guess that’s “small”, right? Oh, and your excuse about cyclone Tracy? More obfuscation. Why don’t you check to see if there was a big adjustment in 1974-75 as they were rebuilding before you throw in crap? Cause guess what, Nick . . . there is no big adjustment at that time. In fact, most of the 1 deg C of adjustments occur before that time.

    Giorgio’s analysis still has nothing at all to do with whether the adjustments at Darwin are right, Nick. Willis claimed that these adjustments to Darwin look suspicious. They do. That has nothing to do with whether this is widespread – or merely limited to Darwin. Thanks for trying to throw that back in there, though . . . and without addressing my previous criticisms, to boot.

  34. Ryan O said

    Ken:

    Ryan O, my only point with the NS post was about the seeming intent of GISS, and I believe also GHCN, in dismissing local (and importantly adjusted) temperatures for use by researchers and instead using the smear called regional and global temperatures.

    Oh, I know what you’re saying . . . but Nick’s quote isn’t the best way to get that across. Especially now that he’s deliberately being blind to the obvious implication that the resultant series no longer best represents the local temperature and is, as you say, a smear.

  35. Nick Stokes said

    Ken #32
    You can get BoM adjustment data here. If you look in the adjustment file, they list max and min adjustments separately. Some are for one year only, which I have excluded. Here are the min adjustments:
    14015 1021 1991 0 -0.3 -0.3 dm detect move
    14015 1021 1987 0 -0.3 -0.6 dm* detect move documentation unclear
    14015 1021 1964 0 -0.6 -1.2 orm* objective test time change move documentation unclear
    14015 1021 1942 0 -1.0 -2.2 oda objective test detect composite move
    14015 1021 1894 0 +0.3 -1.9 fds median detect stevenson screen supplied

    Here are the max adjustments, which don’t all correspond in time:
    14015 1001 1982 0 -0.5 -0.5 or
    14015 1001 1967 0 +0.5 +0.0 or
    14015 1001 1942 0 -0.6 -0.6 da
    14015 1001 1907 0 -0.3 -0.9 rd
    14015 1001 1894 0 -1.0 -1.9 rds

    1942 is presumably the move to Marrara. 1964/7 is probably connected with the move to civil terminal. Somewhat to my surprise, they don’t seem to have a change soon after 1974.

    Remember, the GHCN adjustments are based on the time series only (break point recognition), and the years won’t correspond exactly. I don’t know why BoM adjustments for max and min don’t coincide more in time.

  36. dribble said

    Nick Stokes quoting from scientist Blair: “The post-1941 adjustments (all small) at Darwin Airport relate to a number of site moves within the airport boundary. These days it’s on the opposite side to the terminal, not too far from the Stuart Highway.”

    If you look at the BOM homogenization data for Darwin Airport available as part of the BOM’s supposed High Quality dataset, you find that the mean temperatures for the airport have been adjusted down 0.6 – 0.7 degrees from 1942 to 1980. In order to make the splice at 1942 the Darwin Post Office data has therefore been adjusted down a whopping 1.4 degrees. This downshift in the past has the effect of increasing the mean Darwin Airport trend of 0.76 degrees/century from 1943 to 2008 to 1.3 degrees/century for the total homogenized record 1910 to 2008. Are these post-1941 adjustments small? Don’t think so.

  37. Ryan O said

    And, as per usual, when called on his BS, Nick simply moves on as if nothing ever happened . . .

  38. kdk33 said

    I looked briefly at stations in Colorado. I didn’t notice anything nefarious in the adjustments – in fact, the homogenized data showed less warming than WRCC data (that I take to be raw data). There is a lot of variability in the data. I’m skeptical that one can pull a signal from this noise that doesn’t have large error bars. I’m now more curious about error bounds on estimates derived from surface instrument data sets. Brief description below…

    My ‘survey’ wasn’t scientific – I just chose random stations with long temperature histories. I pulled data from the WRCC site and from the NASA GISS site. From the GISS site I used the ‘RAW + UHSCN ad.’ and the ‘homogenized’ data. The lines below list the station and the trends (deg C / century) per the WRCC data, the ‘RAW + UHSCN ad.’ data, and then the GISS data.

    Grand Junction: 0.61, 0.47, 0.61
    Fraser: 0.04, 0.11, 0.11
    Silverton: 0.54, -0.35, -0.35
    Durango: 0.34, -0.69, 0.43
    Lamar: -0.8, 0.56, 0.5
    Steamboat Spring: 0.98, 0.86, 0.54,
    Wray: 0.4, -0.8, 0.55
    Montrose #2: 1.01, 2.1, 0.47
    Cheyenne Wells: 1.11, 1.12, 0.49

    When I take the average and standard deviation of the trends from the three data sets I get (data set, average, sd)

    WRCC: 0.47, 0.59
    GISS (raw + UHSCN ad): 0.46, 0.84
    GISS (homogenized: 0.37, 0.31

    So the homogenized data has less of a warming trend (0.37 deg/cent) than does the WRCC data (0.47 deg/dent) – at least for the stations I selected, assuming no math errors.

    I am struck by the amount of adjustment. Also, none of the averages are statistically different than zero at the 0.95 confidence level.

    So, as stated above, I’m tending to think less of an artifially introduced trend, than an overstated certainty in the estimated (corrected, homogenized, etc) trend.

    …I could be wrong.

  39. Kenneth Fritsch said

    In this post I wanted to document my statements from my previous post in this thread.

    My posts on differences between Version 1 and Version 2 for individual USHCN station temperatures are contained in the thread which is linked using either of these links below:

    http://climateaudit.org/2009/05/12/5982/

    http://www.climateaudit.org/?p=5982

    Some histograms are linked here:

    It is important to note that Version 1 was for years the final word by USHCN and presented as a well documented and reliable data set – yet when we choose a 1920 to present time period we find the mean US lower 48 is significantly different between series versions. Some individual stations show very large changes from Version 1 to Version 2. I find it interesting that the authors of the paper describing these changes appear to be avoiding talking about these differences. I would assume they are “protecting” the image of certainty, or at least a level of certainty, here.

    Below I link to the portion of trend in the lower 48 US temperatures due to adjustments:

    http://www.ncdc.noaa.gov/oa/climate/research/ushcn/ushcn.html

  40. Ryan O said

    #38 I agree that is the most likely outcome – overstated certainty rather than overstated trend. Probably a little of the latter, but the biggest problem is the former. I believe Ken has often made the same observation.

    #39 Didn’t you find the difference between V1 and V2 to get larger and more positive for V2 as you moved forward from ~1940 or so?

  41. Nick Stokes said

    Dribble #36.
    I’ve been doing some calculations using an R program which calculates trend adjustments for the whole GHCN dataset. For Darwin 0, I get 2.35 C/century, which is indeed a lot. But when you look at the whole picture, out of 6736 stations, the mean adjustment to trend was just 0.175C/century, std dev 1.89. Darwin ranked 576 out of that 6753. It’s in the tail.

  42. Kenneth Fritsch said

    Ryan O at Post #40:

    I went back to the CA post (linked directly below) where I posted the graphs with difference breakpoints for V2-V1 and V2-GISS. I found that the graphs are no longer there or linked. I suspect this all happened in the recent changeover at CA.

    http://www.climateaudit.org/?p=6598#comment-350803

    I have linked these graphs and do indeed show, when the straight line segments are highlighted by breakpoint analysis, that coming forward in time the differences between temperature series yield steeper trends. I really think that breakpoint analysis could be a major tool in some of these studies with temperatures, and particularly so where one can see that a straight line regression through a bramble patch of data is not showing or explaining anything over sub-periods of time within the time series displayed.

    Also as part of that CA post, I had taken the cases with the greatest changes in trends for individual USHCN stations between V1 and V2 and then run a breakpoint analysis on the “adjusted” V2 data. The point behind this treatment was to determine whether the breakpoints were reduced to those that we would expect to see for the contiguous US. That was not the case, leaving the alternative conclusion/conjectures that the adjustments did not remove all non-homogeneities or that climate is very local with its own set of changes independent of the larger region.

    I continue to be puzzled by what climate scientists do not study or do study only superficially.

  43. Ryan O said

    Ah, thanks for reposting. 😉 I must think about these.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
%d bloggers like this: